Method for detecting risky driving behavior based on multimodal feature fusion

Abstract

The invention discloses a method for detecting risky driving behavior based on multimodal feature fusion, belonging to the technical field of driving behavior detection and vehicle safety. The method aims to solve the problems of strong environmental sensitivity, insufficient feature representation and poor robustness of single-modal detection technologies. The key points of its technical scheme are: synchronously collecting multimodal data including driver’s visual data and vehicle dynamic data; performing preprocessing and feature extraction to obtain a visual feature sequence and a vehicle dynamic feature sequence; subjecting the two to multi-level feature fusion including at least one of early fusion, mid-term fusion or late fusion to obtain a comprehensive feature representation; finally, judging the type of risky driving behavior through a classification model based on the comprehensive feature representation and executing an early warning. The invention improves detection accuracy and system robustness, and realizes effective identification of behaviors such as fatigue driving and distracted driving.